Make the claim easy to challenge.

The AUR research loop

A method is useful when another careful researcher can understand what was asked, what information was available, what would count against the claim, and which constraints still matter. The sequence below is a public explanation of those standards. It does not expose active implementation details.

1. 01

   Ask whether information can exist before price

   Start with a mechanism, not a pattern. What could be observed before the price response, and why would that observation be available to the research system at that moment?

2. 02

   Freeze the hypothesis before checking results

   Write down the question, definitions, timing, scope, and failure conditions before looking at the result that could make the idea feel persuasive. A moving target cannot be tested honestly.

3. 03

   Make causal timing explicit

   Separate when an observation became available from when a decision could have been made. Account for ordering, delays, revisions, and any transformation that could accidentally introduce information from the future.

4. 04

   Try to falsify the claim

   Look for the conditions that should break the proposed mechanism. Challenge definitions, alternative explanations, regime changes, missing data, and results that disappear when reasonable choices are varied.

5. 05

   Use out-of-sample or prospective validation

   Evidence must be tested on information that did not shape the hypothesis. Out-of-sample evaluation and prospective observation are ways to reduce the temptation to treat an in-sample fit as a discovery.

6. 06

   Include realistic costs and operational constraints

   Consider fees, spread, slippage, latency, liquidity, availability, position limits, and the practical work required to observe and act. A result that only works in an idealized environment is not a complete result.

7. 07

   Keep an audit trail

   Preserve the lineage from question to data, transformation, evaluation, and conclusion. Versioned records make it possible to distinguish a planned test from an explanation written after the fact.

8. 08

   Document failures and publish only public-safe lessons

   Failed ideas are part of the research output. AUR aims to share what a failure teaches about method and reasoning while keeping sensitive parameters, active signals, and private infrastructure out of the public record.

What this standard means in practice

A fluent AI-generated explanation can make a weak hypothesis feel inevitable. The research system therefore has to create friction: explicit definitions, fixed questions, timing checks, adversarial review, and evidence that is harder to optimize toward. AI can help widen the search and sharpen the challenge, but it does not lower the standard for proof.

The method is deliberately compatible with uncertainty. A result may be interesting without being durable. A failed test may be valuable without being publishable in full. Public research should leave readers with a more accurate map of what was tried, what can be said, and what remains unknown.